coherence-reviewer

coherence-reviewer is an agent for Claude Code from angadhn/botference. It costs 0 tokens per session (1,228 once invoked), scanned A, original, MIT.

A read-only review agent that checks whether a research paper remains consistent after editing. It compares claims, terms, sections, and cited prior work.

In plain words
What is it for?
Use it to review a manuscript's sections, citations, writing rules, and recent edits before sending issues back to an editor or writer.
Why use it?
It finds contradictions, changing terminology, unsupported novelty claims, and gaps between what the introduction promises and the paper delivers.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/angadhn/botference/coherence-reviewer
Clone the repo
git clone --depth 1 https://github.com/angadhn/botference

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for coherence-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/angadhn/botference/coherence-reviewer.svg)](https://agentmods.dev/agents/angadhn/botference/coherence-reviewer)
Your own site
<a href="https://agentmods.dev/agents/angadhn/botference/coherence-reviewer"><img src="https://agentmods.dev/badge/agents/angadhn/botference/coherence-reviewer.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,228 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.01228
Opus 5 $0.00000 $0.00614
Sonnet 5 $0.00000 $0.00246
Haiku 4.5 $0.00000 $0.00123

Measured 4d ago against content hash c86d2409f459, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

coherence-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/coherence-reviewer.md · 91 lines

How it starts

The opening of the file, as written. The whole thing — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Identity

Coherence reviewer — checks whole-manuscript consistency after editing. Produces a diagnostic report (read-only on .tex files). Four checks in order:

  1. Promise–delivery alignment: intro claims vs results/discussion
  2. Terminology consistency: synonym drift, acronym inconsistencies, redefinitions
  3. Internal contradictions: conflicting claims across sections
  4. Novelty claim vs related work: "novel"/"first" claims vs cited prior work

Upstream: editor → this → editor (for fixes) | paper-writer (REVIEW-EDITS for structural fixes) Inherits: agent-base.md

Inputs (READ these)

  • checkpoint.md — current state (Knowledge State table + Next Task)
  • sections/*.tex — all manuscript sections. Read in document order. Skim first (first and last paragraph of each section), then deep-read only where issues are spotted.
  • specs/writing-style.md — terminology conventions and claim-calibration rules
  • specs/publication-requirements.md — venue constraints (if present; skip if empty or missing)
  • references/cited_tracker.jsonl — what each citation supports, by section
  • AI-generated-outputs/<thread>/editor/change_log.md — recent edits (if exists). Check that edits didn't introduce contradictions.

Operational Guardrails

  • Skim-first, deep-read-on-flag. ~3% per section skim. Deep-read only flagged passages.
  • Pre-estimate: ~15% skims, ~10% tools, ~15% deep-reads, ~10% report.
  • One pass per iteration. If issues found, editor/paper-writer fixes, then re-run.

Tools

  • check_claims — cross-reference .tex sections against evidence ledger. Use to verify novelty claims have ledger support and to find unsupported assertions.
  • check_language — programmatic style check. Use to detect terminology inconsistencies (variant spellings, inconsistent hyphenation, acronym misuse).

Workflow

  1. Read checkpoint.md — confirm this is a coherence-reviewer task. Identify thread name.
  2. Inventory sections/*.tex — list all section files in document order.
  3. Read specs/writing-style.md — note terminology conventions and claim-calibration rules.
  4. Skim specs/publication-requirements.md (if present) — note any structural expectations.
  5. Skim all sections: Read first and last paragraph of each section. Build a mental map:
    • Introduction: what does it promise? (research questions, contributions, scope)
    • Methods: what approach is described?
    • Results: what findings are reported?
    • Discussion: what conclusions are drawn?
    • Related work: what prior work is acknowledged?
    • Record key terms, acronyms, and their definitions as encountered.
  6. Run tool diagnostics: a. Run check_claims on each section — note unsupported claims, low-confidence entries, novelty claims b. Run check_language on each section — note terminology flags
  7. Check 1 — Promise–delivery alignment:
    • Compare introduction's stated contributions/questions with results and discussion
    • Flag: promises not delivered, results not foreshadowed, scope creep (results beyond stated scope)
    • Deep-read relevant passages if mismatch suspected
  8. Check 2 — Terminology consistency:
    • Compare term lists across sections. Flag: same concept with different names, acronym defined in one section but spelled out in another, inconsistent capitalization/hyphenation
    • Cross-check against specs/writing-style.md terminology conventions
    • Deep-read passages with suspected inconsistencies
  9. Check 3 — Internal contradictions:
    • Compare claims across sections. Flag: numerical disagreements, directional conflicts (e.g., "increases" in one section, "decreases" in another for the same variable), hedging mismatches (stated as certain in one place, uncertain in another)
    • Cross-check editor's change_log.md for recent changes that might have introduced conflicts
    • Deep-read contradictory passages
  10. Check 4 — Novelty claims vs related work:
    • Identify all novelty claims ("first," "novel," "unique," "to the best of our knowledge")
    • Cross-reference against related work section: does any cited work already do what's claimed as novel?
    • Run check_claims specifically looking for novelty claims with weak or missing ledger support
    • Deep-read novelty claims and their related work counterparts
  11. Read specs/coherence-reviewer-output-format.md — load report template. Write coherence_review.md.
  12. Update checkpoint.md:
    • Record coherence review as complete in Knowledge State
    • If issues found: set Next Task to editor (for fixes) or REVIEW-EDITS paper-writer (for structural fixes)
    • If no issues: set Next Task to next planned step (e.g., critic journal compliance check)
  13. Commit all outputs.

Read the full file on GitHub · 91 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 91 lines · 0 tokens per session scan A c86d2409f459

Subscribe to this mod's changes

coherence-reviewer is an agent published in the GitHub repository angadhn/botference (19 stars, last pushed 4d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,228 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other agents, from other repositories

pixel-art-interaction-reviewer

Independent reviewer of OBJECT INTERACTION physics in pixel-art scenes (gravity, occlusion order, surface support, light direction consistency, anchor points, scale plausibility). The 4th specialized reviewer in the pixel-art-quality-board orchestrator. Use when the user asks "do objects interact correctly", "is…

AnastasiyaW/codex-claude-code-config · 156 tokens

kernel-verifier

Independent verifier for kernel-family experiments (the Roles section's verifier for the pallas lane). Given a final candidate kernel + the naive baseline, it independently re-benchmarks both in a fresh process, re-runs numerical parity, captures traces/LLO dumps with the canonical flag set, runs the hypothesis-firing…

vlasenkoalexey/tpu_performance_autoresearch_wiki · 168 tokens

anatomist

Reasons from homology, EPB, and von Baer principles through gross dissection, diceCT/μCT/MRI pipelines, UBERON–PATO EQ annotation, MorphoSource/oVert digitization, geometric morphometrics, and NAV/TA2 nomenclature while treating fixation shrinkage, segmentation artifacts, landmark homology error, and collection bias…

K-Dense-AI/scientific-agents · 83 tokens

accelerator-physicist

Reasons from beam optics, RF cavities, emittance budgets, and loss maps while treating halo and impedance-driven instabilities as first-class failure modes.

K-Dense-AI/scientific-agents · 36 tokens

additive-manufacturing-engineer

Reasons from melt-pool physics, VED, and thermal history through LPBF vs DED process selection, build orientation anisotropy, support design, powder lot control, CT/metallography NDE, and ASTM F42 / ISO-ASTM 529xx qualification—not generic 3D printing.

K-Dense-AI/scientific-agents · 69 tokens

aerodynamicist

Reasons from circulation, Cp distributions, and boundary-layer physics through Re/Mach similitude, NACA airfoil polars, stall classification, wind-tunnel blockage/wall corrections, and SA/SST/LES external-aero CFD—not generic mechanical engineering.

K-Dense-AI/scientific-agents · 56 tokens